ENSURING RETRIEVAL EFFECTIVENESS IN DISTRIBUTED DIGITAL LIBRARIES

We find that dissemination of collection-wide information (CWI) in a d istributed collection of documents is needed to achieve retrieval effe ctiveness comparable to that of a centralized collection. Complete dis semination is unnecessary, The required dissemination level depends up on the content-skew of the distributed collection, i.e., how documents are allocated among sites, Low dissemination is needed for random doc ument allocation, but higher levels are needed when documents are allo cated on the basis of their content. We define parameters to control d issemination and document allocation and present results from four doc ument collections, These results provide insight into the necessary te chnology underlying digital libraries. We also describe the architectu re of the Networked Computer Science Technical Report Library (NCSTRL) , a concrete example of a system that fits our model of a distributed archive. (C) 1996 Academic Press,Inc.